The ~$99B Revenue BacklogNarrow moat

CoreWeave (CRWV) — moat facet

Years of committed revenue, on paper — the number the valuation leans on.

The headline evidence for CoreWeave's durability is its revenue backlog, which reached roughly $99 billion by early 2026, having grown from around $67 billion1 just months before. By June 2026 the filed figure, remaining performance obligations, was $103.7 billion, of which 41% is expected to be recognised within 24 months.2 Against annual revenue still in the single-digit billions, a backlog approaching $100 billion is extraordinary — it implies that years of rapid growth are already contracted, and it is the single strongest argument that CoreWeave is building a real, durable business rather than merely riding a temporary compute shortage. The backlog underpins everything: it gives investors revenue visibility, it supports the debt financing that funds the GPU fleet, and it is the foundation of the bull case that CoreWeave will convert commitments into profitable scale.

When the June 2026 RPO is expected to turn into revenue ($ billion)Within 24 months$42.5B (41%)Months 25 to 48$40.4B (39%)Months 49 to 78$20.7B (20%)CoreWeave 10-Q Q2 2026, RPO note; dollar amounts from the stated percentages of $103.7B
Less than half of the order book turns into revenue within two years; the rest depends on counterparties staying solvent to 2032.

The backlog is genuinely impressive and a real asset. But the honest and essential caveat, developed in the accompanying threat, is that backlog is a promise, not revenue, and certainly not profit. It must be converted over years, which requires CoreWeave to build the enormous, expensive capacity to fulfill it — itself a capital and execution challenge — and it depends on the counterparties remaining willing and able to pay. A $104 billion backlog tells you demand has been contracted; it does not guarantee the revenue will arrive on schedule, at the expected margins, or without renegotiation if the market shifts. The very speed of the backlog's growth, while thrilling, also raises the question of whether the revenue ramp — and the capacity, capital, and profitability to serve it — can actually keep pace. The backlog is the best reason to believe in CoreWeave's future, a real and enormous book of committed business; but an investor should hold it as a promise to be executed against real risks, not as revenue in hand, and should watch the conversion of backlog into profitable revenue as the truest test of whether the commitments become the business the valuation assumes.

Moat trajectory: Widening

Widening fast — the backlog grew from ~$67B to ~$99B in months, implying years of committed growth. But it's a promise that must be executed (capacity, capital, margin risk), not revenue in hand — the ramp must catch up.

The number that tests this moat
Reported
RPO growth in six months
$103.7B at June 2026, from $60.7B at end-2025

The filed backlog measure rose 71% in six months; it must convert at the rate the RPO schedule says, 41% within two years, or the growth is on paper.

Source: CoreWeave 10-Q Q2 2026 and 10-K FY2025 (RPO notes) ↗
⚠ Threats to the moat
References
  1. ReportedBacklog ~$99B by early 2026, from ~$67B just months before.
    CoreWeave Q1/Q2 2026 earnings releases — Q2 revenue ~$2.5B (+111%); revenue backlog $99.4B (Mar 2026, from $66.8B at end-2025); 2026 capex guided $31–35B; ~$28B of financing raised in 12 months; quarterly interest expense >$500M; ten clients >$1B each — Q1-Q2 2026 · publ. 2026 · source ↗
  2. ReportedBy June 2026 remaining performance obligations were $103.7 billion, 41% expected within 24 months.
    CoreWeave Form 10-Q for the quarter ended 30 June 2026 — revenue $2,575M (Q2) and $4,653M (H1); interest expense, net $640M; depreciation on property and equipment $1.4B; net loss $626M; RPO $103.7B (41% within 24 months, 39% in months 25-48; $60.7B at 31 December 2025 per the FY2025 10-K); Customer A 36%, B 26%, C 10% of Q2 revenue (71% a year earlier); capex $14,117M in H1; gross property and equipment $52,622M — Q2 2026 · publ. 2026-08-12 · source ↗
Sources
Generated September 23, 2026